Research On Remote Sensing Image Registration Based On Multi-feature Fusion And Populationbased Incremental Learning Algorithm

Image registration technique is an important technology in image processing technique,it can achieve the matching between two or more images with offset.Image registration technique is widely used in many fields,such as change detection,image fusion,image mosaic,3D modeling,computer vision,and so on.In many image processing techniques based on image registration,the effect of image registration has a direct impact on the effect of the subsequent.Therefore,to improve the registration effect of image registration has always been our goal,in order to pursue this goal,this paper proposes a technical scheme based on Scale-invariant feature transform(SIFT)and mutual information(MI)for image registration.This paper first describes the research background,research status and development direction of image processing and image registration techniques,and analysis the mathematical concepts and principles of image registration,and illustrates several common methods of registration and the corresponding model.Which of this methods,this paper emphatically introduces the matching algorithm based on SIFT,this algorithm extracts the feature of image,this feature with the scale invariance and rotation invariance,which has strong adaptability to the influence of illumination,noise and so on,and has a relatively high discrimination.Then this paper introduces the principle and the mathematical definition of mutual information,and then proposes a search algorithm –PBIL algorithm,which based on SIFT and mutual information.Taking mutual information as similarity measure,the PBIL algorithm is used to search the feature points on the basis of SIFT to find the optimal transformation parameters,so as to improve the matching accuracy.Finally,the multi-scale Harris feature is added on the set of the SIFT feature,and two features are matched by using dual matching,then set a condition to filter the feature,and features satisfying the condition in the two features are put into a set.Taking mutual information as similarity measure,the PBIL algorithm is used to search the feature points to find the optimal transformation parameters,and improve the matching accuracy.